Image classification it generally requires a priori knowledge about the
objects to be classified. In this paper, we present a new method to segment
tumor in multispectral magnetic resonance (MR) images of the human brain.
The proposed approach, called Minimum Variance Distortionless Response
beamforming (MVDR) was introduced in [15] where only the knowledge of
the desired signature to be classified was required. It was a special case of
Linearly Constrained Minimum Variance Beamforming (LCMV) in array
processing. MVDR considers an MR image classification problem as an
array-processing problem where each sensor represents one spectral band. It
uses a finite impulse response (FIR) filter to minimize the output power
while the desired signature is constrained to a specific gain. That is the
response of the beamformer is constrained to equal unity at the electrical
angle. The method has been evaluated through several experiments. Results
show that the cerebral tissue was segmented accurately into four images,
tumor, gray matter, white matter and cerebral spinal fluid indicating the
possible usefulness of this method. As far as computing saving is concerned,
the experimental results also show computational complexity improvement.
